【作者】 董娜； 王繁珍； 陈增强； 孙青林； 袁著祉；

【Author】 Dong Na Wang Fanzhen Chen Zengqiang Sun Qinglin Yuan Zhuzhi (College of Information Technique Science Nankai Unversity Tianjin 300071 China)

【机构】 南开大学信息技术科学学院；

【摘要】 粒子群算法是一种全局优化算法。用粒子群算法对BP网络进行训练,可以克服BP算法的一些缺陷。然而粒子群优化算法(PSO)的缺点在于:对于复杂的多峰函数,容易陷入到局部极值点中,得不到正确结果。此外,由于缺乏精密搜索方法的配合,PSO方法往往不能得到精确的结果。本文针对PSO的上述缺点,使用了一种新的基于Tent映射混沌粒子群优化算法(TCPSO)来进行神经网络训练并将其应用到神经网络预测控制的滚动优化中,用于解决带有约束的控制问题,最后通过在典型的非线性系统上所做的仿真结果,说明了此种算法的良好性能。更多还原

【Abstract】 Particle Swarm Optimization is a global optimization algorithm.Training BP network by PSO can overcome some shortcomings in BP network.However,PSO also has its own shortcomings.For complicated multimodal function,PSO can be easily trapped to local minima and can hardly get correct results.Moreover,as PSO lacks of the support of precise searching method,it is hard to generate correct results.Aiming at PSO’S defects,this paper uses a new tent-map chaotic particle swarm optimization(TCPSO) to train BP network and applies this method into Neural Network-based Predictive Control(NNPC),solving the control problems with constraints.At last,this method is tested on a typical nonlinear plant,through which the effectiveness of the proposed control method is fully illustrated.更多还原